A complete workflow for conducting PRISMA-compliant systematic reviews using nanopublications for transparent, traceable, and machine-readable research documentation.
This pipeline implements a 5-stage PRISMA workflow:
- PICO Research Question - Define your research question
- Search Strategy - Document your search methodology
- Search Execution - Query academic databases via APIs
- Study Screening - AI-assisted title/abstract screening with ASReview
- Study Assessment - Document eligibility criteria and methodology
Each stage produces nanopublications that create a complete provenance chain from research question to included studies.
# Create conda environment
conda create -n systematic-review python=3.11
conda activate systematic-review
# Install required packages
pip install jupyter pandas numpy requests
pip install rdflib nanopub
pip install asreview asreview-insights# Check ASReview
asreview --version
# Check nanopub
python -c "from nanopub import Nanopub; print('nanopub OK')"Notebook: pico-nanopub-from-json.ipynb
Generates a nanopublication defining your research question using the PICO framework (Population, Intervention, Comparison, Outcome).
Input: pico-config.json
Output: PICO nanopub (e.g., https://w3id.org/np/RA...)
Notebook: search-strategy-nanopub-from-json.ipynb
Documents your search methodology including databases, search terms, Boolean operators, and filters.
Input: search-strategy-config.json
Output: Search Strategy nanopub
Notebook: search-execution-api-queries.ipynb
Programmatically queries academic databases via their APIs:
- OpenAlex
- arXiv
- PubMed (E-utilities)
- Europe PMC
- Semantic Scholar
| File | Format | Purpose |
|---|---|---|
search_results_combined.csv |
CSV | Data analysis, deduplication |
search_results_combined.ris |
RIS | Import to screening tools (Rayyan, Zotero, Covidence) |
search_results_combined.bib |
BibTeX | Citation management (LaTeX, JabRef) |
search_summary.json |
JSON | Machine-readable search metadata |
Nanopub Notebook: search-execution-nanopub-from-json.ipynb
AI-assisted title/abstract screening using active learning.
# Start the web interface
asreview lab
# Opens http://localhost:5001 in your browser- Create Project: Click FILE → Upload your
search_results_combined.ris - Start Screening: Go to Reviewer tab
- Make Decisions: Click "Relevant" or "Irrelevant" for each paper
- AI Learning: After ~10-20 decisions, ASReview prioritizes likely-relevant papers
- Stop Screening: When you see mostly irrelevant papers (AI has found most relevant ones)
- Export Results: FILE → Export project as
.asreviewfile
- You always make the final decision (Relevant/Irrelevant)
- AI learns from your decisions and prioritizes which paper to show next
- Active Learning means relevant papers are shown first, saving screening time
- Typical efficiency: Review ~30-40% of papers to find >95% of relevant studies
| Code | Meaning |
|---|---|
offtopic_quantum |
No quantum computing content |
offtopic_biodiversity |
No biodiversity application |
wrong_type_review |
Review/editorial only |
wrong_type_abstract |
Conference abstract only |
duplicate |
Duplicate study |
language |
Not in English |
unavailable |
Full text not accessible |
retracted |
Retracted paper |
Notebook: asreview-to-nanopub.ipynb
Extracts screening decisions from .asreview file and generates:
study_inclusion.json- Ready for nanopub generationincluded_studies.csv/.ris- Included papersexcluded_studies.csv/.ris- Excluded papersprisma_flow_data.json- PRISMA diagram numbers
Notebook: study-inclusion-nanopub-asreview.ipynb
Generates ONE nanopub per included study, linking each to your systematic review.
Input: study_inclusion.json (from ASReview export)
Output: Multiple .trig files (one per study with DOI)
Notebook: study-assessment-nanopub-from-json.ipynb
Generates ONE nanopub summarizing your entire assessment methodology:
- Eligibility criteria (PRISMA Item 5)
- Assessment technique (PRISMA Item 11)
- Study characteristics (PRISMA Item 17)
- Quality assessment (PRISMA Item 18)
- Study results (PRISMA Item 19)
Input: study-assessment-config.json
Output: Study Assessment nanopub
| Stage | Nanopub URI |
|---|---|
| PICO Research Question | https://w3id.org/np/RA8B3ptXUOsN7obpkFGtA0FBmsh0OnID53wOsUIpSKTcg |
| Search Strategy | https://w3id.org/np/RAJW9kn9Syx7y_1Okl4HPwqUlUssxi0daadJNM1AT8-PU |
| Search Execution | https://w3id.org/np/RAMPy96eCLCXlGR9VvCVf6rJmpN_DlxxarMGm91_5n-O8 |
| Study Inclusion (238 nanopubs) | See published_uris.json |
| Study Assessment Dataset | https://w3id.org/np/RAx_ZQScbvsz7Rvqk8scSYx06zojCc6Gjcvkxjj_MKwVM |
All screening data deposited at: https://zenodo.org/records/18070378
| Metric | Count |
|---|---|
| Total records | 1,649 |
| Records screened | 569 |
| Included | 283 |
| Excluded | 286 |
| Not screened (AI stopped) | 1,080 |
| Nanopubs published | 238 |
├── pico-nanopub-from-json.ipynb
├── search-strategy-nanopub-from-json.ipynb
├── search-execution-api-queries.ipynb
├── search-execution-nanopub-from-json.ipynb
├── asreview-to-nanopub.ipynb
├── study-inclusion-nanopub-asreview.ipynb
├── study-assessment-nanopub-from-json.ipynb
└── README.md
├── pico-config-template.json
├── search-strategy-config-template.json
├── search-execution-config-template.json
└── study-assessment-config-template.json
# Large data files (deposit in Zenodo instead)
*.csv
*.ris
*.bib
*.asreview
# Generated nanopubs
*.trig
*.signed.trig
# Output directories
screening_results/
nanopubs_study_inclusion/
# Personal config with real data
*-config.json
!*-template.json
# Python
__pycache__/
*.pyc
.ipynb_checkpoints/
# Data files (deposit in Zenodo)
*.csv
*.ris
*.bib
*.asreview
search_summary.json
# Generated nanopubs
*.trig
# Output directories
screening_results/
nanopubs_study_inclusion/
# Jupyter checkpoints
.ipynb_checkpoints/
# Python cache
__pycache__/
*.pyc
# Config files with real data (keep templates)
study_inclusion.json
published_uris.json┌─────────────────────────────────────────────────────────────────┐
│ SYSTEMATIC REVIEW PIPELINE │
├─────────────────────────────────────────────────────────────────┤
│ │
│ 1. PICO Question ──► Nanopub │
│ │ │
│ ▼ │
│ 2. Search Strategy ──► Nanopub │
│ │ │
│ ▼ │
│ 3. Search Execution ──► .ris/.csv files ──► Nanopub │
│ │ │
│ ▼ │
│ 4. ASReview Screening ──► .asreview project │
│ │ │
│ ▼ │
│ 5a. Study Inclusion ──► 1 Nanopub per study │
│ │ │
│ ▼ │
│ 5b. Study Assessment ──► 1 Summary Nanopub │
│ │ │
│ ▼ │
│ Zenodo: Deposit all data files │
│ │
└─────────────────────────────────────────────────────────────────┘
CC-BY 4.0
Anne Fouilloux (ORCID: 0000-0002-1784-2920)